1
Determinants of the Status of an International Financial Centre
Imad Moosa, Larry Li and Riley Jiang
School of Economics, Finance and Marketing
445 Swanston Street
Melbourne 3000, Victoria
Australia
Abstract
To identify the determinants of the status of an international financial centre we consider 24 potential factors and the technique of extreme bounds analysis (EBA). By identifying three free variables we apply EBA to the remaining 21 variables, running a total of 3990 regressions, 190 regressions for each of the variables of interest. The results of conventional EBA reveal two robust variables only, and the same finding is obtained by using restricted EBA. The estimates of the fraction of the cumulative distribution function falling on one side of zero reveal eight important determinants of the status of an international financial centre, including the two robust variables identified by using EBA.
Introduction
An international financial centre is characterised by agglomeration of financial institutions providing financial services on an international level. According to Zhao et al. (2004), an international financial centre “refers to a global city that provides a full spectrum of high-end services, as financial services cannot be independent of other specialized services”. In March 2013, data were released on the ranking of international financial centres as measured by the global financial centre index (Z/Yen, 2013). This index (GFCI) provides profiles, ratings and rankings of financial centres from around the world. The index is calculated based on “external measures” and responses to an onlinesurvey completed by international financial services professionals. Respondents are asked to rate those centres that they are familiar with and to answer a number of questions pertaining to their perceptions of competitiveness.
The GFCI report was first published by the Z/Yen Group in March 2007 and has subsequently been updated every six months. According to the latest GFCI report (March 2013), London, New York, Hong Kong andSingapore remain the top four centres—it seems that London has maintained its number one rank despite the LIBOR scandal. At the bottom, we find Reykjavik, Budapest and Athens, with Athens being 68points adrift of Budapest. Figure 1 displays the top ten (in descending order) and bottom ten (in ascending order) financial centres measured in terms of the GFCI. A question arises as to what makes London the “best” and Athens the “worst” international financial centre.
The objective of this study is to identify the factors that determine the status of aninternational financial centre as measured by the GFCI. Specifically this is a cross-sectional study whereby the GFCI is modelled in terms of economic, financial and regulatory factors (24 of them). In a cross-sectional study like this, we are certain to encounter the problem of sensitivity of the estimated coefficients with respect to model specification because no single theoretical model can be used to identify an explicit set of explanatory variables to be included in any empirical model. The consequence is that researchers find it tantalising to try various combinations of the explanatory variables and report the ones they like, typically the ones that produce “good” results and/or confirm pre-conceived beliefs.
This study is based on the technique of extreme bounds analysis (EBA), as suggested by Leamer (1983, 1985) and the extensions proposed by Granger and Uhlig (1991) and by Sala-I-Martin (1997). By using this technique, we test the robustness of coefficient estimates to changes in the conditioning set of information (represented by the explanatory variables). EBA allows us to avoid the problem of selecting the combinations of explanatory variables to appear in the “optimal model”. Furthermore, the procedure circumvents the problems of data mining and bias in the reporting of results. We will report the results obtained by using conventional EBA, restricted EBA and the cumulative distribution function (CDF).
Literature Review
The literature on international financial centres typically deals with issues pertaining to the factors that can be used to identify a financial centre, why there is spatial agglomeration of financial activity, and why financial services are spatially concentrated in selected locations. The basic underlying question is why financial services remain embedded within international financial centres when technology would seem to facilitate deconcenration and geographical dispersion (Faulconbridge, 2004). This question falls under the general issue of enterprise location(where and why firms place specific activities in particular areas), which is a key area of interest in both international business research (for example, Alcacer and Chung, 2007; Nachum and Wymbs, 2005; Porter, 2001) and economic geography research (for example, Krugman, 1991; Lorenzen and Mudambi, 2013; Markusen, 1996). Some scholars express the view that despite rising interest in location, our current understanding of the geographic aspect of multinational enterprise (MNE) behaviour remains underdeveloped (McCann, 2011; Ricart et al., 2004).
Many scholars believe that the economic role of space has become increasingly insignificant, in the sense that location of the enterprise does not matter in the age of electronic communications and electronic money (for example, Castellas, 1989; O’Brien, 1992; Cairncross, 1997; Ohame, 1990, 1995a, 1995b; Kobrin, 1997). However, there are those who believe the opposite—that spatial proximity is still critical because not all types of information can be transmitted over distance with constant costs (for example, Berry et al., 1997, Sassen, 1995; Short and Kim, 1999). According to Cantwell (2009), the revival of interest in the locational concentration or dispersion of activity can be attributed in large part to the paradox between the apparent death of distance, as pointed out by Cairncross (1997), and the renewed significance of local clusters that are poles of attraction to innovation and entrepreneurship, as stressed by Breschi and Malerba (2001). Cantwell attributes the resurgence of interest in issues of the location of international business to the pioneering work of Dunning (1998).
Goerzen et al. (2013) combine the concept of location derived by economic geographers with theories of the multinational enterprise (MNE) and the liability of foreignness developed by international business scholars to examine the factors that propel MNEs towards, or away from, “global cities”. They argue that three distinctive characteristics of global cities (global interconnectedness, cosmopolitanism, and abundance of advanced producer services) help MNEs overcome the costs of doing business abroad. Based on a multilevel multinomial model, their analysis of a large sample of MNE location decisions suggests not only that MNEs have a strong propensity to locate within global cities,but also that these choices are associated with a nuanced interplay of firm (andsubsidiary-level) factors, including investment motives, proprietary capabilitiesand business strategy.
The development of international financial centres offers a good example to illustrate the ongoing significance of enterprise location.A characteristic of the literature on international financial centres is that there is no consensus on a single definition of an international financial centre, or even just a financial centre. A financial centre is defined by Reed (1998) as a “place where providers of and customers for financial services meet to transact business”. According to Lai (2006) a financial centre is a “conglomeration of financial and service enterprises and corporate headquarters, particularly foreign ones”. Jao (1997) defines an international financial centre as a “place in which there is a high concentration of banks and other financial institutions, and in which a comprehensive set of financial markets are allowed to exist and develop, so that financial activities and transactions can be effectuated more efficiently than at any other locality”. Fakitesi (2009) lists some of the characteristics of an international financial centre as follows: (i) it is conducive to the conduct of international financial business profitably, easily and efficiently; (ii) there is abundance of skilled management and intellectual talent covering business, finance and interdependent services; (iii) it offers deep liquid and sophisticated capital markets and world competitive tax and regulatory regimes with foreign investment and offshore business flow; (iv) it can add significant value to financial services through a workforce that can respond promptly and in an innovative manner; (v) it offers high quality telecommunications and IT capacity as well as well educated, multilingual workforce; (vi) all facets of financial services can be located efficiently; and (vii) it provides a convivial and alluring environment for business. Park (2011) distinguishes between an international financial centre and a domestic one on the following grounds: (i) international centres deal in various major currencies of the world, not just the currency of the country where a centre is located; (ii) most of the financial transactions conducted in foreign currencies in international centres are generally free of taxes and exchange controls; and (iii) international financial centres provide various financial services to both resident and non-resident clients.
Several studies have been conducted on what makes a particular city an international financial centre. By using London as an example, Thrift (1994) concludes that international financial centres have a particular set of locational determinants, arguing that local characteristics and localised information jointly define the advantages of a given location as a financial centre. Wojcik (2009) concludes on the basis of previous research that “an important part of information used in financial markets is not easily transferable across space, resulting in the significance of local financial relations and spatial concentration of financial firms”.Martin (1999) asserts that thefriction of information flows across physical distance affects the location of financial activities, as informationcollection and verification are particularly crucial forfinancial business. Leyshon (1995,1997) emphasises the political-economicapproach to the formation of geographiesof money and finance, arguing that a wide range of social factors might contribute tothe survival and success of international financial centresin particular places.Zhao et al. (2004) attempt to explain why foreign financial services are spatially concentrated in a particular city to form an international financial centre. By examining various forces behind the formation of a financial centre, they argue that “information problems have created the necessity of the geographic agglomeration of financial activities in the source of information even in the era when financial markets have worked through sophisticated telecommunication networks”.
Kerr (1965), Park and Essayyad (1989) and Porteous (1999) suggest that severalmeasurescan be used to identifyspecific cities that function as financial centres.These measures include employment in the financialsector relative to total employment, assets of financialinstitutions, the proportion of cheques cashed, theturnover value of stock exchange, the volume of communications(particularly express mail and telecommunications),and the presence of foreign banks and headoffices of large multinational non-financial corporations.Kayral et al. (2012) use a logistic regression and the GFIC (as the dependent variable) to find positive relationswith the efficiency and strength of the legal rights, the variability of the labour force participation rate, and the underlying centres being in the top20% group.
Porteous (1995) has proposed a theoretical frameworkto find out why financial activities tend toagglomerate in one particular location rather thananother. His framework emphasises the key roleof information flow (with respect to informationaccessibility and reliability) in influencing the locationof financial activities. He focuses ontwo information concepts for their effect on thedevelopment of a financial centre: information hinterlandand information asymmetry. The informationhinterland is defined as the region for which a particularcore city, acting as that regional centre, providesthe best access point for the profitable exploitation ofvaluable information flows. The effect of information asymmetry is to push financial firms closer to an information source in order to find and interpret non-standardised information that a financial firm can use to make profit.
Some studies deal with the current and expected (or hoped for) status of particular financial centres. Sarigul (2012) uses SWOT analysis to show that Istanbul has significant strengths tobecome an international financial centre. In particular he stresses that Istanbul is “well with its geographicalposition, young and increasingly educated labour force, no numerical shortage inthe labour market, high quality and skilled labour force in the field of banking, twomodern airports that are well connected to the financial business centres andother major global cities”. Roberts (2008) attributes the status ofLondon to the developments inBritain in thescope of world economics, explaining how Londonappeared as the first financial centre and how it kept its place by adapting to new economic conditions. For example, it was at one time thought that the launch of the euro and the location of the European Central Bank in Frankfurt could form a threat to the status of London. This, however, has not happened (Faulconbridge, 2004). Schenk (2002) attributes the status of Hong Kong as an international financial centre to the transparency andeffectiveness of financial services. Seade (2009) examinesEast Asian financial centresand concludes that Hong Kong, Tokyo, Shanghai and Singapore are about to catch up with their most powerful competitors, London and New York. Giap (2009) studies the progress of Singapore and appraises its rise in Asia. According to Shirai (2009) and Kawai (2009), although Japan is one of the most economically powerful countries inthe world, Tokyo remains behind New York and London because of “its unsatisfactory trading volume”.Mingqi (2009) and Bhattacharya (2010) attribute the status of Shanghai to economic reform in China.
The literature, therefore, does not provide a limited set of the factors that determine the status of an international financial centre. A large number of potential explanatory variables must be considered in the absence of a theoretical model and the availability of a diverse set of hypotheses emphasising different factors. This is why a straightforward cross-sectional analysis is likely to produce subjective results that exhibit confirmation and publication bias. This is also why the methodology described in the following section is chosen to conduct the analysis.
Methodology
Cross-sectional studies are typically based on a cross-sectional regression of the form:
(1)
wherethe dependent variable y is explained in terms of n explanatory variables,’s. These studies invariably report a sample of regressions encompassing various combinations of the explanatory variables. The reported regressions are chosen for convenience because they vindicate the researcher’s pre-conceived notions. This problem arises because the theory is not adequately explicit about what variables should appear in the “true” model. For example, the following situation is often encountered: may be significant when the regression includes and, but not when is included. So, which combination of all available ’s do we choose?
Extreme bounds analysis can be used to find out if there is robustness in the determinants of the dependent variable. Hussain and Brookins (2001) argue that the usual practice of reporting a preferred model with its diagnostic tests need not be sufficient to convey the degree of reliability of the determinants (explanatory variables). By calculating upper and lower bounds for the parameter of interest from all possible combinations of potential explanatory variables, it is possible to assess and report the sensitivity of the estimated coefficients to specification changes. The relation between the dependent variable and a given explanatory variable is considered to be robust if the estimated coefficient on that variable remains statistically significant without a change of sign when the set of explanatory variables are changed.
EBA is based on a linear regression of the form
(2)
where is an explanatory variable that is always included in the regression because its importance has been established by previous studies (and because it makes sense theoretically or even intuitively), Q is the variable whose robustness is under consideration, and is a potentially important variable. The ’s are called “free variables”, whereas Q is called the “variable of interest”. According to Sala-I-Martin (1997), the free variables “need to be ‘good’ a priori”, in the sense that “they have to be widely used in the literature” and that they are “somewhat ‘robust’ in the sense that they systematically seem to matter in all regressions in the previous literature”. Otherwise, the choice of the free variables may be justified on the basis of theoretical and intuitive considerations.
The procedure involves varying the set of Z variables included in the regression to find the widest range of coefficients on the variable of interest, , that standard tests of significance do not reject. If the extreme values remain significant and of the same sign, then one can infer that the result (and hence, the variable of interest) is “robust”. Otherwise, the variable is “fragile”. In other words, for a variable of interest to be robust, and must be significant and of the same sign. This is equivalent to reporting and , where is the standard deviation of for , where m is the number of estimated regressions for each variable of interest. In this case if the minimum value of is negative and the maximum value of is positive, the variable of interest is not robust.
A typical finding of studies using these criteria to determine whether a variable is robust or fragile is that very few (or no) variables are robust (for example, Levine and Renelt, 1992). While the results can be interpreted to imply that the variable under consideration are not robust, Sala-I-Martin suggests that the test is too strong for any variable to pass. He argues that “if the distribution of has some positive and some negative support, then one is bound to find one regression for which the estimated coefficient changes signs if enough regressions are run”. It is for this reason that a number of attempts have been made to refine the robustness criteria in order to reduce the probability of obtaining “unreasonable” extreme bounds. As a result, a “reasonable” EBA test has been developed to estimate the extreme bounds on the coefficient of interest by eliminating models with poor goodness of fit as measured by. Granger and Uhling (1990) proposed this refinement of EBA by imposing a condition on the level of goodness of fit, such that all models with low are irrelevant for the calculation of extreme bounds. This criterion is represented by